did 6 Codility tests in 3 months and kept notes on every single proctoring-report flag, here's the actual list
"similarity to public solution." someone on this sub got that on a Codility back in November and the recruiter rejected them at the report stage without even calling. that is the only flag i have personally seen end a candidacy. medium leetcode-flavored problem, they recognized it from their grind months back, pasted the old answer in. instant kill. and it is the one flag literally nobody on reddit talks about. every thread i can find is people guessing about the keystroke stuff or the gaze tracker.
i started keeping a paper notebook after that thread, because i had six Codility tests lined up over the next three months. backend mid-level, fintech, the layoff drills got too frequent so i bailed in October. all six were live, no take-home. two FAANG-adjacent shops, a trading firm, a healthcare giant, two early-stage startups. every flag the report threw, what i was doing when it threw, what the recruiter said about it on the followup if anything was said. that's what this post is.
so. similarity check first since it is the one that actually kills you. it is deterministic, structural diff against a corpus of public answers. i rewrote everything from scratch on all six tests even on stuff i could solve in my sleep. zero similarity flags. the time penalty was way smaller than i was scared it would be. ten minutes maybe across a 90 minute test. that one rule alone would have saved that sub poster and they knew the problem cold lol.
now everything else. external display detection fires immediately. my second monitor lit up the report on test one and the recruiter called the next day, super casual, asking if i had been looking at notes. i told her no i just forgot to unplug. she laughed, sent me to the next round. so the flag fires but nobody acts on it without context. tab switching gets logged with timestamps, which i confirmed by tabbing out twice on test two (water break, then dog losing his mind at the mailman) and the recruiter mentioning the tab events on the followup. i told her about the dog, she laughed, fine. so two flags on the record across two tests and zero negative consequences from either.
test three is where i got the actual intel. i asked the recruiter at the end of the loop what was literally in the report and she pulled it up on zoom and walked me through every page of it. paste events get logged with character counts so a 200 char paste of a function looks really different from a 20 char paste of a method signature, and the report shows BOTH numbers. they don't see what you pasted, just the size and the timestamp. copy events get logged but they don't track where you pasted it. tab switches with timestamps. focus loss. webcam (when enabled) does posture and gaze checks and there is a "looking away" counter on the report. she told me she literally never looks at the gaze stuff unless something else on the report is already lit up first.
the typing rhythm thing that everyone on twitter loses their mind about, "keystroke pattern variance", that's in the report. i asked her about it directly. she said nobody on her team looks at it. asked the test five recruiter the same question and got the same answer. so the data point exists but nobody is actually consuming it. dead weight on the report unless other flags are stacked on top of it. tbh i went into test five expecting a different answer and it kind of surprised me to hear the same one twice.
what is NOT in the report at all, and i went through every one of mine and what that sub poster had on theirs too. no process scan. no list of running applications. no DNS or network monitoring of any kind. no clipboard contents, just the event count. and no detection at all of anything happening outside the browser tab itself. browser sandbox. it watches the tab, it watches the webcam if you let it. that is the entire surface area of what the system can see. i asked the trading firm recruiter that exact question on test four and she said yeah Codility just hands them a browser session, that's all there is to it.
empirical TLDR. don't paste big blocks of code (obvious). don't tab out a million times (a few is fine, the recruiter doesn't care unless something else is already wrong). don't submit verbatim leetcode answers because that is the one flag that actually kills the candidacy. and anything running outside the browser tab itself is invisible to the report because the report isn't looking there. that's not advice, that's just what six reports said.
anyone else keeping notes on what these proctoring reports actually flag versus what we imagine they flag? would love more sample sizes from HackerRank and CoderPad which i haven't done as many of.